The irreducible epistemic atoms underlying the curriculum. 4,828 atoms across 8 types and 2 molecules
Stopping rule: a pre-defined criterion or condition that determines when an optimization effort should cease, consisting of three components: a threshold (what constitutes sufficient improvement), a metric (how to measure progress toward that threshold), and a trigger (what action to take when the threshold is met).
Randomization: the process of randomly assigning inputs to different versions in an A/B test to distribute confounding variables evenly across groups so that any systematic difference in outcomes can be attributed to the treatment rather than background conditions
Confounding: the technical term for what goes wrong when you change multiple variables simultaneously, where a confounding variable correlates with both your change and your outcome, making it impossible to determine which one actually caused the result
Variable isolation: the discipline of changing one thing at a time so that observed effects can be attributed to specific causes
Ablation study: a formalization of variable isolation in machine learning that systematically removes or disables individual components of a model and measures how performance changes to determine each component's contribution
Local optimum: a performance configuration within a fitness landscape that represents the peak of improvement potential within the current framework, where further incremental changes yield no measurable improvement because all directions are downhill.
Framework: the set of unquestioned assumptions, architectural principles, and operational constraints that define the boundaries within which a personal agent operates, determining what actions feel available and what features appear relevant to the agent's performance.
Kaizen: the disciplined practice of continuous, incremental improvement within an existing framework, characterized by small, measurable gains that compound over time to produce significant cumulative value.
Kaikaku: the radical reform of an existing framework through wholesale replacement, characterized by high-stakes, rare events that produce substantial improvements (30-50%) and establish a new baseline for subsequent kaizen.
Speed optimization: the systematic process of reducing the time and friction required for a personal agent to execute, thereby increasing execution frequency and enabling the agent to operate at lower motivation thresholds while maintaining or improving output quality
Overhead: the non-execution time in an agent's operation that includes setup, navigation, context-switching, waiting, and searching, which does not directly contribute to the agent's output but is necessary for the process
Accuracy: the degree to which an agent's output corresponds to the intended outcome, measured by the hit rate of actions taken and decomposed into bias (systematic error) and noise (random error)
Reliability optimization: the process of redesigning agents to fire consistently under varying conditions by building structural support mechanisms such as fallback paths, backup triggers, and minimum viable execution modes
Scope optimization: the practice of expanding or narrowing the situations and actions an agent handles until the agent's boundary matches its purpose
Scope creep: the gradual, uncontrolled expansion of project or agent scope beyond its original boundaries that leads to increased costs, delays, and reduced effectiveness
Energy optimization: the discipline of achieving results with minimal expenditure — cognitive, emotional, or physical
Emotional energy: the cost of feeling — managing anxiety, suppressing frustration, performing social roles, and processing interpersonal dynamics
Physical energy: the cost of doing — maintaining posture, managing fatigue, regulating sleep debt, and sustaining the metabolic processes that fuel both cognition and emotion
Integration tax: the cumulative cost of all transitions, handoffs, translations, and context switches between agents in a system, which can exceed the execution cost of the individual components
Integration optimization: the practice of optimizing how agents connect and hand off to each other, rather than optimizing each agent's performance in isolation, with the goal of reducing unnecessary friction at integration points while maintaining flexibility
Transition cost: the cognitive or operational cost incurred when one process hands off to another, including time lost in reconfiguration, information degradation, and energy spent on context switching
Subtractive optimization: the discipline of making a process better by removing parts of it rather than adding improvements, where the fastest step is the one that does not exist and the most reliable component is the one that was never built
Addition bias: the systematic human tendency to default to adding elements when asked to improve or change something, even when removing elements would be simpler, cheaper, and more effective, with this bias being stronger under cognitive load and when not explicitly cued
Via negativa: the approach of defining something by what it is not, rather than by what it is, particularly in decision-making where subtractive knowledge is more robust than additive knowledge because we can be more confident about what harms us than what helps us